Opened, Clicked, or Actually Read? Three Different Things

The three events

They happen in sequence, and each one can be the last. Naming them separately is most of the work, because the vocabulary the tools use has collapsed all three into the word “view”.

Event 1 — the click. Somebody, or something, followed the link. This is the earliest and cheapest signal and it is the one web analytics is built around. Google Analytics is precise about what its download event actually means, and the precision is worth quoting because it is so much narrower than how the number gets used:

“when a user clicks a link leading to a file (with a common file extension)”— Google, Analytics Help: enhanced measurement events (file_download)

A click on a link. Not an open, not a page, not a second of anybody's attention. The same caveat applies to any “opened” notification that rests on a fetch rather than on a rendering session — the mechanics of that are on how page-by-page PDF analytics actually work.

Event 2 — the open. The document rendered somewhere. This is a genuine step up from a click, because something had to load and display, but it still says nothing about a person. It is also the event most likely to be manufactured: corporate mail security follows links before anybody sees the message, and Microsoft documents Safe Links as scanning URLs prior to message delivery. Apple describes the same artefact from the consumer side, where remote content is

“privately downloaded in the background when you receive a message (instead of when you view it)”— Apple, Mail Privacy Protection support documentation

Event 3 — the read. Pages turned, in an order, over time. This is the only one of the three that produces a shape rather than a boolean, and it is the only one that can be wrong in the reader's favour as well as against them — a slow, thorough reader and a distracted one look different, where a click looks identical either way. Sharing a document as a tracked link is what puts you in a position to see the third event at all, and it is free.

What the split looks like in real data

We ran the split on our own production analytics, because the question “how big is the gap” has no published answer anywhere else we could find. Every recorded session on a multi-page document was sorted by the furthest page it ever rendered: none at all, page one only, or something beyond page one.

Two things about that table of three are worth saying plainly, because both cut against the number being as dramatic as it looks. First, it counts every recorded session, automated ones included — that is deliberate, since the group which renders no page at all disappears entirely under any bot filter and the point of the exercise is to keep it visible.

Second, and more importantly if you have read our other numbers: this page says 11.5% rendered no page, where whether PDF view counts are real publishes 14.4% for what looks like the same thing. Both are right, and the difference is scope rather than disagreement. That page measures the whole corpus; this one measures multi-page documents only, because a one-page document has no second page to get to. Run without the multi-page restriction on the older window, this extract returns exactly the 436 zero-engagement sessions that page publishes — the same number, not a similar one. Apply the restriction to that same window and it is 340 of 2,884, or 11.8%. Dropping the documents outside that scope — one-page ones, and those with no page count on record — takes sessions off both sides of that fraction — the zero-page count falls from 436 to 340 along with the total, which is why the rate moves down rather than up. Extend the window from there to today and 11.8% becomes the 11.5% this page reports: three more weeks of reading, not a third change of scope.

The middle group is the one worth sitting with. A session that renders page one and stops is not a bot and not a mistake; it is a real person who opened the thing and decided within about fourteen seconds that it was not for them. That is the single most useful verdict a sender can have, and it is precisely the verdict a view count destroys by folding it in with the readers. How long the sessions that do continue actually run is a separate question with its own data, on how long people spend reading a PDF.

What each signal can and cannot measure

The axis that matters is not accuracy. It is when the measurement stops — because every one of these mechanisms is still perfectly correct about the moment it captured, and simply has nothing to say about the moment after it.

Read the third column and the table collapses into one sentence: a signal that fires once can only ever answer a question about that one instant. Two rows deserve their due rather than a dismissal. Google Analytics is genuinely the right tool for a PDF published openly on a website — it will tell you which page drove the click, which a document tracker cannot, and it simply stops at the click. And DRM wins a row outright: it can tell you the document was printed, and we cannot. Printing happens in software we do not control, so if paper leaving the building is the question, that is the category to buy — at the cost of your reader installing a viewer before they can read anything.

Why one number hides all three

Because a counter has to increment on something, and the earliest available event is the cheapest one to build on. That is an observation about the mechanism, not about anyone's motives: the earlier the trigger fires, the larger the number it produces, and the less of that number had a person behind it. Nobody has to be dishonest for this to happen. A metric named “views” that increments on a click is accurate about clicks and misleading about reading, and both of those are true at once.

The inflation is not evenly distributed either, which is what makes it hard to correct for by instinct. Automated traffic exceeded half of all web traffic in Imperva's 2026 report, and it does not skim — a scanner follows the link, loads what loads, and leaves. Its footprint sits entirely inside events one and two, so the earlier your counter fires, the larger the share of it that was never a person. That is the whole argument for measuring later rather than earlier, and it is why page-by-page reading data sits on our free plan rather than behind the paywall the category usually puts it behind.

How to tell them apart

Share the document as a link to a hosted copy rather than as a file, and the distinction stops being a philosophical one. The page doing the serving is still yours, so it can see which page is on screen and for how long — not because it is cleverer than a pixel, but because it is still in the room. Three questions become answerable in order: did anything render, did it get past page one, and where did it stop. You can see the shape of that on the live dashboard running on sample data, without an account.

It is worth being exact about where this stops, because the limit is permanent and it is the same for every tool in the table. Measurement ends at the download. Once a copy is on somebody else's disk it can be read ten times or never, and nothing here can tell you which — a product claiming otherwise is describing DRM or overstating. What a hosted link changes is not the ceiling. It is everything below it.

Free is not a trial here, and it is worth saying where it stops. Per-page reading time, the drop-off page, the email gate, password, link expiry and download control are all on it, along with 50 files, 50 live links and twelve months of history — and so is the per-person verdict on your first document, which is the thing this page is about. Pro, at $9 a month, is what you buy when the second document needs the same answer as the first: the per-recipient verdict across everything you send rather than one document, return-visit and cross-document comparison, more than 50 files or 50 links live at once, and an alert the moment something opens rather than in tomorrow morning's digest. The trigger is how many documents you are running, never the reading data itself — that is never taken away from you.

Keep reading: how page-by-page PDF analytics work, step by step, and how free PDF tracking works, with the limits stated.